Tridimensional image of the extraocular muscles deformations in Graves' orbitopathy: implications of partial volume effects.

Author(s)

André Domingos Araújo Souza

• Document type: Doctoral Thesis
• Press: Ribeirão Preto.
• Institution: Universidade de São Paulo (USP). Faculdade de Filosofia, Ciências e Letras de Ribeirão Preto (PCARP/BC)
• Defense date: 2002-03-22

Examining board members

Evandro Eduardo Seron Ruiz; Antonio Augusto Velasco e Cruz; Alexandre Xavier Falcão; Roberto de Alencar Lotufo; Agma Juci Machado Traina

• Advisor: Evandro Eduardo Seron Ruiz

Abstract

The extraocular muscles (EOM), which are responsible for the eyes movements, are presented enlarged in their dimensions in Graves’ orbitopathy. These deformations can lead patients to blindness. In clinical routine, physicians normally evaluate, in computer tomography (CT) images, the diameter of the EOM by manual tracing to check if they are enlarged. However, the accuracy of the EOM measurements is impaired by the subjectivity of these manual methods. Further, the time consuming is also one of the main drawbacks on these methods. This way we present an EOM segmentation method (MSEG) that overcomes the difficulties pointed above. The MSEG method is based on the Laplacian-of-Gaussian operator (LOG) combined with the mathematical morphology theory. We have taken into account the effect of discretization and numerical truncation during the LOG implementation. In CT, partial volume effects (PVE) cause several artifacts in volume rendering. In order to create 3D rendition more reliable to carry out anatomical measures and also to pursue superior quality of display of both soft-tissue and bone, we introduce two methods for detecting and classifying voxels with PVE (MCLA) based on a new approach. A method is described to automatically peel skin so that PVE-resolved renditions of bone and soft-tissue reveal considerably more details. We have conducted experiments to evaluate quantitatively and qualitatively all methods proposed here. The MSEG method is well correlated with manual tracing in our experiments (R=0,92). Surface renditions are created from EOM CT dataset segmented using the MSEG method. We have also conducted a quantitative evaluation in patients with Graves’ orbitopathy wherein the EOM volume ratio in the orbit (TO) was T=34,3%, which is about five times higher than in normal patient (TO=7,3%). Volume renditions have been created before and after applying the methods for several patient CT datasets. A mathematical phantom experiment involving different levels of PVE has been conducted by adding different degrees of noise and blurring. A quantitative evaluation was performed using the mathematical phantom and clinical CT data wherein an operator carefully masked out voxels with PVE in the segmented images. All results have demonstrated the enhanced quality of display of bone and soft tissue after applying the proposed methods. The quantitative evaluations indicate that more than 98% of the voxels with PVE are removed by the two methods and the second method performs slightly better than the first. Further, skin peeling vividly reveals fine details in the soft tissue structures. 3D renditions should be used with care in radiology in view of artifacts demonstrated in this work coming from PVE. Finally, we have estimated volume errors in the EOM models higher than 25% if PVE is not properly handled. (AU)